The dkan_data module provides the connection to Drupal as a means of storage.
A dataset is an identifiable collection of structured data objects (distribution list) unified by some criteria (authorship, subject, scope, spatial or temporal extent…) as well as metadata.
DKAN adds a content type called data. This content type will hold the metadata of a dataset in JSON format. Additional data nodes will be created for specific sub-elements of the dataset as well, (i.e. publisher, theme, keywords, and distribution). These data nodes will provide unique reference ids for the sub-elements and can be accesssed via an API endpoint. Learn more here.
You can customize which sub-elements generate additional data nodes here admin/config/dkan/properties.

When the value of these elements change or become outdated, the corresponding data node will be removed by the orphan_reference_processor queue task.
If you prefer to run it manually, you may do so with: